root / lib / bootstrap.py @ 8a7a197a
History | View | Annotate | Download (41.1 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2006, 2007, 2008, 2010, 2011, 2012 Google Inc.
|
5 |
#
|
6 |
# This program is free software; you can redistribute it and/or modify
|
7 |
# it under the terms of the GNU General Public License as published by
|
8 |
# the Free Software Foundation; either version 2 of the License, or
|
9 |
# (at your option) any later version.
|
10 |
#
|
11 |
# This program is distributed in the hope that it will be useful, but
|
12 |
# WITHOUT ANY WARRANTY; without even the implied warranty of
|
13 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
14 |
# General Public License for more details.
|
15 |
#
|
16 |
# You should have received a copy of the GNU General Public License
|
17 |
# along with this program; if not, write to the Free Software
|
18 |
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
|
19 |
# 02110-1301, USA.
|
20 |
|
21 |
|
22 |
"""Functions to bootstrap a new cluster.
|
23 |
|
24 |
"""
|
25 |
|
26 |
import os |
27 |
import os.path |
28 |
import re |
29 |
import logging |
30 |
import time |
31 |
import tempfile |
32 |
|
33 |
from ganeti.cmdlib import cluster |
34 |
from ganeti import rpc |
35 |
from ganeti import ssh |
36 |
from ganeti import utils |
37 |
from ganeti import errors |
38 |
from ganeti import config |
39 |
from ganeti import constants |
40 |
from ganeti import objects |
41 |
from ganeti import ssconf |
42 |
from ganeti import serializer |
43 |
from ganeti import hypervisor |
44 |
from ganeti.storage import drbd |
45 |
from ganeti.storage import filestorage |
46 |
from ganeti import netutils |
47 |
from ganeti import luxi |
48 |
from ganeti import jstore |
49 |
from ganeti import pathutils |
50 |
|
51 |
|
52 |
# ec_id for InitConfig's temporary reservation manager
|
53 |
_INITCONF_ECID = "initconfig-ecid"
|
54 |
|
55 |
#: After how many seconds daemon must be responsive
|
56 |
_DAEMON_READY_TIMEOUT = 10.0
|
57 |
|
58 |
|
59 |
def _InitSSHSetup(): |
60 |
"""Setup the SSH configuration for the cluster.
|
61 |
|
62 |
This generates a dsa keypair for root, adds the pub key to the
|
63 |
permitted hosts and adds the hostkey to its own known hosts.
|
64 |
|
65 |
"""
|
66 |
priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER) |
67 |
|
68 |
for name in priv_key, pub_key: |
69 |
if os.path.exists(name):
|
70 |
utils.CreateBackup(name) |
71 |
utils.RemoveFile(name) |
72 |
|
73 |
result = utils.RunCmd(["ssh-keygen", "-t", "dsa", |
74 |
"-f", priv_key,
|
75 |
"-q", "-N", ""]) |
76 |
if result.failed:
|
77 |
raise errors.OpExecError("Could not generate ssh keypair, error %s" % |
78 |
result.output) |
79 |
|
80 |
utils.AddAuthorizedKey(auth_keys, utils.ReadFile(pub_key)) |
81 |
|
82 |
|
83 |
def GenerateHmacKey(file_name): |
84 |
"""Writes a new HMAC key.
|
85 |
|
86 |
@type file_name: str
|
87 |
@param file_name: Path to output file
|
88 |
|
89 |
"""
|
90 |
utils.WriteFile(file_name, data="%s\n" % utils.GenerateSecret(), mode=0400, |
91 |
backup=True)
|
92 |
|
93 |
|
94 |
def GenerateClusterCrypto(new_cluster_cert, new_rapi_cert, new_spice_cert, |
95 |
new_confd_hmac_key, new_cds, |
96 |
rapi_cert_pem=None, spice_cert_pem=None, |
97 |
spice_cacert_pem=None, cds=None, |
98 |
nodecert_file=pathutils.NODED_CERT_FILE, |
99 |
rapicert_file=pathutils.RAPI_CERT_FILE, |
100 |
spicecert_file=pathutils.SPICE_CERT_FILE, |
101 |
spicecacert_file=pathutils.SPICE_CACERT_FILE, |
102 |
hmackey_file=pathutils.CONFD_HMAC_KEY, |
103 |
cds_file=pathutils.CLUSTER_DOMAIN_SECRET_FILE): |
104 |
"""Updates the cluster certificates, keys and secrets.
|
105 |
|
106 |
@type new_cluster_cert: bool
|
107 |
@param new_cluster_cert: Whether to generate a new cluster certificate
|
108 |
@type new_rapi_cert: bool
|
109 |
@param new_rapi_cert: Whether to generate a new RAPI certificate
|
110 |
@type new_spice_cert: bool
|
111 |
@param new_spice_cert: Whether to generate a new SPICE certificate
|
112 |
@type new_confd_hmac_key: bool
|
113 |
@param new_confd_hmac_key: Whether to generate a new HMAC key
|
114 |
@type new_cds: bool
|
115 |
@param new_cds: Whether to generate a new cluster domain secret
|
116 |
@type rapi_cert_pem: string
|
117 |
@param rapi_cert_pem: New RAPI certificate in PEM format
|
118 |
@type spice_cert_pem: string
|
119 |
@param spice_cert_pem: New SPICE certificate in PEM format
|
120 |
@type spice_cacert_pem: string
|
121 |
@param spice_cacert_pem: Certificate of the CA that signed the SPICE
|
122 |
certificate, in PEM format
|
123 |
@type cds: string
|
124 |
@param cds: New cluster domain secret
|
125 |
@type nodecert_file: string
|
126 |
@param nodecert_file: optional override of the node cert file path
|
127 |
@type rapicert_file: string
|
128 |
@param rapicert_file: optional override of the rapi cert file path
|
129 |
@type spicecert_file: string
|
130 |
@param spicecert_file: optional override of the spice cert file path
|
131 |
@type spicecacert_file: string
|
132 |
@param spicecacert_file: optional override of the spice CA cert file path
|
133 |
@type hmackey_file: string
|
134 |
@param hmackey_file: optional override of the hmac key file path
|
135 |
|
136 |
"""
|
137 |
# noded SSL certificate
|
138 |
cluster_cert_exists = os.path.exists(nodecert_file) |
139 |
if new_cluster_cert or not cluster_cert_exists: |
140 |
if cluster_cert_exists:
|
141 |
utils.CreateBackup(nodecert_file) |
142 |
|
143 |
logging.debug("Generating new cluster certificate at %s", nodecert_file)
|
144 |
utils.GenerateSelfSignedSslCert(nodecert_file) |
145 |
|
146 |
# confd HMAC key
|
147 |
if new_confd_hmac_key or not os.path.exists(hmackey_file): |
148 |
logging.debug("Writing new confd HMAC key to %s", hmackey_file)
|
149 |
GenerateHmacKey(hmackey_file) |
150 |
|
151 |
# RAPI
|
152 |
rapi_cert_exists = os.path.exists(rapicert_file) |
153 |
|
154 |
if rapi_cert_pem:
|
155 |
# Assume rapi_pem contains a valid PEM-formatted certificate and key
|
156 |
logging.debug("Writing RAPI certificate at %s", rapicert_file)
|
157 |
utils.WriteFile(rapicert_file, data=rapi_cert_pem, backup=True)
|
158 |
|
159 |
elif new_rapi_cert or not rapi_cert_exists: |
160 |
if rapi_cert_exists:
|
161 |
utils.CreateBackup(rapicert_file) |
162 |
|
163 |
logging.debug("Generating new RAPI certificate at %s", rapicert_file)
|
164 |
utils.GenerateSelfSignedSslCert(rapicert_file) |
165 |
|
166 |
# SPICE
|
167 |
spice_cert_exists = os.path.exists(spicecert_file) |
168 |
spice_cacert_exists = os.path.exists(spicecacert_file) |
169 |
if spice_cert_pem:
|
170 |
# spice_cert_pem implies also spice_cacert_pem
|
171 |
logging.debug("Writing SPICE certificate at %s", spicecert_file)
|
172 |
utils.WriteFile(spicecert_file, data=spice_cert_pem, backup=True)
|
173 |
logging.debug("Writing SPICE CA certificate at %s", spicecacert_file)
|
174 |
utils.WriteFile(spicecacert_file, data=spice_cacert_pem, backup=True)
|
175 |
elif new_spice_cert or not spice_cert_exists: |
176 |
if spice_cert_exists:
|
177 |
utils.CreateBackup(spicecert_file) |
178 |
if spice_cacert_exists:
|
179 |
utils.CreateBackup(spicecacert_file) |
180 |
|
181 |
logging.debug("Generating new self-signed SPICE certificate at %s",
|
182 |
spicecert_file) |
183 |
(_, cert_pem) = utils.GenerateSelfSignedSslCert(spicecert_file) |
184 |
|
185 |
# Self-signed certificate -> the public certificate is also the CA public
|
186 |
# certificate
|
187 |
logging.debug("Writing the public certificate to %s",
|
188 |
spicecert_file) |
189 |
utils.io.WriteFile(spicecacert_file, mode=0400, data=cert_pem)
|
190 |
|
191 |
# Cluster domain secret
|
192 |
if cds:
|
193 |
logging.debug("Writing cluster domain secret to %s", cds_file)
|
194 |
utils.WriteFile(cds_file, data=cds, backup=True)
|
195 |
|
196 |
elif new_cds or not os.path.exists(cds_file): |
197 |
logging.debug("Generating new cluster domain secret at %s", cds_file)
|
198 |
GenerateHmacKey(cds_file) |
199 |
|
200 |
|
201 |
def _InitGanetiServerSetup(master_name): |
202 |
"""Setup the necessary configuration for the initial node daemon.
|
203 |
|
204 |
This creates the nodepass file containing the shared password for
|
205 |
the cluster, generates the SSL certificate and starts the node daemon.
|
206 |
|
207 |
@type master_name: str
|
208 |
@param master_name: Name of the master node
|
209 |
|
210 |
"""
|
211 |
# Generate cluster secrets
|
212 |
GenerateClusterCrypto(True, False, False, False, False) |
213 |
|
214 |
result = utils.RunCmd([pathutils.DAEMON_UTIL, "start", constants.NODED])
|
215 |
if result.failed:
|
216 |
raise errors.OpExecError("Could not start the node daemon, command %s" |
217 |
" had exitcode %s and error %s" %
|
218 |
(result.cmd, result.exit_code, result.output)) |
219 |
|
220 |
_WaitForNodeDaemon(master_name) |
221 |
|
222 |
|
223 |
def _WaitForNodeDaemon(node_name): |
224 |
"""Wait for node daemon to become responsive.
|
225 |
|
226 |
"""
|
227 |
def _CheckNodeDaemon(): |
228 |
# Pylint bug <http://www.logilab.org/ticket/35642>
|
229 |
# pylint: disable=E1101
|
230 |
result = rpc.BootstrapRunner().call_version([node_name])[node_name] |
231 |
if result.fail_msg:
|
232 |
raise utils.RetryAgain()
|
233 |
|
234 |
try:
|
235 |
utils.Retry(_CheckNodeDaemon, 1.0, _DAEMON_READY_TIMEOUT)
|
236 |
except utils.RetryTimeout:
|
237 |
raise errors.OpExecError("Node daemon on %s didn't answer queries within" |
238 |
" %s seconds" % (node_name, _DAEMON_READY_TIMEOUT))
|
239 |
|
240 |
|
241 |
def _WaitForMasterDaemon(): |
242 |
"""Wait for master daemon to become responsive.
|
243 |
|
244 |
"""
|
245 |
def _CheckMasterDaemon(): |
246 |
try:
|
247 |
cl = luxi.Client() |
248 |
(cluster_name, ) = cl.QueryConfigValues(["cluster_name"])
|
249 |
except Exception: |
250 |
raise utils.RetryAgain()
|
251 |
|
252 |
logging.debug("Received cluster name %s from master", cluster_name)
|
253 |
|
254 |
try:
|
255 |
utils.Retry(_CheckMasterDaemon, 1.0, _DAEMON_READY_TIMEOUT)
|
256 |
except utils.RetryTimeout:
|
257 |
raise errors.OpExecError("Master daemon didn't answer queries within" |
258 |
" %s seconds" % _DAEMON_READY_TIMEOUT)
|
259 |
|
260 |
|
261 |
def _WaitForSshDaemon(hostname, port, family): |
262 |
"""Wait for SSH daemon to become responsive.
|
263 |
|
264 |
"""
|
265 |
hostip = netutils.GetHostname(name=hostname, family=family).ip |
266 |
|
267 |
def _CheckSshDaemon(): |
268 |
if netutils.TcpPing(hostip, port, timeout=1.0, live_port_needed=True): |
269 |
logging.debug("SSH daemon on %s:%s (IP address %s) has become"
|
270 |
" responsive", hostname, port, hostip)
|
271 |
else:
|
272 |
raise utils.RetryAgain()
|
273 |
|
274 |
try:
|
275 |
utils.Retry(_CheckSshDaemon, 1.0, _DAEMON_READY_TIMEOUT)
|
276 |
except utils.RetryTimeout:
|
277 |
raise errors.OpExecError("SSH daemon on %s:%s (IP address %s) didn't" |
278 |
" become responsive within %s seconds" %
|
279 |
(hostname, port, hostip, _DAEMON_READY_TIMEOUT)) |
280 |
|
281 |
|
282 |
def RunNodeSetupCmd(cluster_name, node, basecmd, debug, verbose, |
283 |
use_cluster_key, ask_key, strict_host_check, data): |
284 |
"""Runs a command to configure something on a remote machine.
|
285 |
|
286 |
@type cluster_name: string
|
287 |
@param cluster_name: Cluster name
|
288 |
@type node: string
|
289 |
@param node: Node name
|
290 |
@type basecmd: string
|
291 |
@param basecmd: Base command (path on the remote machine)
|
292 |
@type debug: bool
|
293 |
@param debug: Enable debug output
|
294 |
@type verbose: bool
|
295 |
@param verbose: Enable verbose output
|
296 |
@type use_cluster_key: bool
|
297 |
@param use_cluster_key: See L{ssh.SshRunner.BuildCmd}
|
298 |
@type ask_key: bool
|
299 |
@param ask_key: See L{ssh.SshRunner.BuildCmd}
|
300 |
@type strict_host_check: bool
|
301 |
@param strict_host_check: See L{ssh.SshRunner.BuildCmd}
|
302 |
@param data: JSON-serializable input data for script (passed to stdin)
|
303 |
|
304 |
"""
|
305 |
cmd = [basecmd] |
306 |
|
307 |
# Pass --debug/--verbose to the external script if set on our invocation
|
308 |
if debug:
|
309 |
cmd.append("--debug")
|
310 |
|
311 |
if verbose:
|
312 |
cmd.append("--verbose")
|
313 |
|
314 |
family = ssconf.SimpleStore().GetPrimaryIPFamily() |
315 |
srun = ssh.SshRunner(cluster_name, |
316 |
ipv6=(family == netutils.IP6Address.family)) |
317 |
scmd = srun.BuildCmd(node, constants.SSH_LOGIN_USER, |
318 |
utils.ShellQuoteArgs(cmd), |
319 |
batch=False, ask_key=ask_key, quiet=False, |
320 |
strict_host_check=strict_host_check, |
321 |
use_cluster_key=use_cluster_key) |
322 |
|
323 |
tempfh = tempfile.TemporaryFile() |
324 |
try:
|
325 |
tempfh.write(serializer.DumpJson(data)) |
326 |
tempfh.seek(0)
|
327 |
|
328 |
result = utils.RunCmd(scmd, interactive=True, input_fd=tempfh)
|
329 |
finally:
|
330 |
tempfh.close() |
331 |
|
332 |
if result.failed:
|
333 |
raise errors.OpExecError("Command '%s' failed: %s" % |
334 |
(result.cmd, result.fail_reason)) |
335 |
|
336 |
_WaitForSshDaemon(node, netutils.GetDaemonPort(constants.SSH), family) |
337 |
|
338 |
|
339 |
def _InitFileStorageDir(file_storage_dir): |
340 |
"""Initialize if needed the file storage.
|
341 |
|
342 |
@param file_storage_dir: the user-supplied value
|
343 |
@return: either empty string (if file storage was disabled at build
|
344 |
time) or the normalized path to the storage directory
|
345 |
|
346 |
"""
|
347 |
file_storage_dir = os.path.normpath(file_storage_dir) |
348 |
|
349 |
if not os.path.isabs(file_storage_dir): |
350 |
raise errors.OpPrereqError("File storage directory '%s' is not an absolute" |
351 |
" path" % file_storage_dir, errors.ECODE_INVAL)
|
352 |
|
353 |
if not os.path.exists(file_storage_dir): |
354 |
try:
|
355 |
os.makedirs(file_storage_dir, 0750)
|
356 |
except OSError, err: |
357 |
raise errors.OpPrereqError("Cannot create file storage directory" |
358 |
" '%s': %s" % (file_storage_dir, err),
|
359 |
errors.ECODE_ENVIRON) |
360 |
|
361 |
if not os.path.isdir(file_storage_dir): |
362 |
raise errors.OpPrereqError("The file storage directory '%s' is not" |
363 |
" a directory." % file_storage_dir,
|
364 |
errors.ECODE_ENVIRON) |
365 |
|
366 |
return file_storage_dir
|
367 |
|
368 |
|
369 |
def _PrepareFileBasedStorage( |
370 |
enabled_disk_templates, file_storage_dir, |
371 |
default_dir, file_disk_template, |
372 |
init_fn=_InitFileStorageDir, acceptance_fn=None):
|
373 |
"""Checks if a file-base storage type is enabled and inits the dir.
|
374 |
|
375 |
@type enabled_disk_templates: list of string
|
376 |
@param enabled_disk_templates: list of enabled disk templates
|
377 |
@type file_storage_dir: string
|
378 |
@param file_storage_dir: the file storage directory
|
379 |
@type default_dir: string
|
380 |
@param default_dir: default file storage directory when C{file_storage_dir}
|
381 |
is 'None'
|
382 |
@type file_disk_template: string
|
383 |
@param file_disk_template: a disk template whose storage type is 'ST_FILE'
|
384 |
@rtype: string
|
385 |
@returns: the name of the actual file storage directory
|
386 |
|
387 |
"""
|
388 |
assert (file_disk_template in |
389 |
utils.storage.GetDiskTemplatesOfStorageType(constants.ST_FILE)) |
390 |
if file_storage_dir is None: |
391 |
file_storage_dir = default_dir |
392 |
if not acceptance_fn: |
393 |
acceptance_fn = \ |
394 |
lambda path: filestorage.CheckFileStoragePathAcceptance(
|
395 |
path, exact_match_ok=True)
|
396 |
|
397 |
cluster.CheckFileStoragePathVsEnabledDiskTemplates( |
398 |
logging.warning, file_storage_dir, enabled_disk_templates) |
399 |
|
400 |
file_storage_enabled = file_disk_template in enabled_disk_templates
|
401 |
if file_storage_enabled:
|
402 |
try:
|
403 |
acceptance_fn(file_storage_dir) |
404 |
except errors.FileStoragePathError as e: |
405 |
raise errors.OpPrereqError(str(e)) |
406 |
result_file_storage_dir = init_fn(file_storage_dir) |
407 |
else:
|
408 |
result_file_storage_dir = file_storage_dir |
409 |
return result_file_storage_dir
|
410 |
|
411 |
|
412 |
def _PrepareFileStorage( |
413 |
enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir, |
414 |
acceptance_fn=None):
|
415 |
"""Checks if file storage is enabled and inits the dir.
|
416 |
|
417 |
@see: C{_PrepareFileBasedStorage}
|
418 |
|
419 |
"""
|
420 |
return _PrepareFileBasedStorage(
|
421 |
enabled_disk_templates, file_storage_dir, |
422 |
pathutils.DEFAULT_FILE_STORAGE_DIR, constants.DT_FILE, |
423 |
init_fn=init_fn, acceptance_fn=acceptance_fn) |
424 |
|
425 |
|
426 |
def _PrepareSharedFileStorage( |
427 |
enabled_disk_templates, file_storage_dir, init_fn=_InitFileStorageDir, |
428 |
acceptance_fn=None):
|
429 |
"""Checks if shared file storage is enabled and inits the dir.
|
430 |
|
431 |
@see: C{_PrepareFileBasedStorage}
|
432 |
|
433 |
"""
|
434 |
return _PrepareFileBasedStorage(
|
435 |
enabled_disk_templates, file_storage_dir, |
436 |
pathutils.DEFAULT_SHARED_FILE_STORAGE_DIR, constants.DT_SHARED_FILE, |
437 |
init_fn=init_fn, acceptance_fn=acceptance_fn) |
438 |
|
439 |
|
440 |
def _InitCheckEnabledDiskTemplates(enabled_disk_templates): |
441 |
"""Checks the sanity of the enabled disk templates.
|
442 |
|
443 |
"""
|
444 |
if not enabled_disk_templates: |
445 |
raise errors.OpPrereqError("Enabled disk templates list must contain at" |
446 |
" least one member", errors.ECODE_INVAL)
|
447 |
invalid_disk_templates = \ |
448 |
set(enabled_disk_templates) - constants.DISK_TEMPLATES
|
449 |
if invalid_disk_templates:
|
450 |
raise errors.OpPrereqError("Enabled disk templates list contains invalid" |
451 |
" entries: %s" % invalid_disk_templates,
|
452 |
errors.ECODE_INVAL) |
453 |
|
454 |
|
455 |
def _RestrictIpolicyToEnabledDiskTemplates(ipolicy, enabled_disk_templates): |
456 |
"""Restricts the ipolicy's disk templates to the enabled ones.
|
457 |
|
458 |
This function clears the ipolicy's list of allowed disk templates from the
|
459 |
ones that are not enabled by the cluster.
|
460 |
|
461 |
@type ipolicy: dict
|
462 |
@param ipolicy: the instance policy
|
463 |
@type enabled_disk_templates: list of string
|
464 |
@param enabled_disk_templates: the list of cluster-wide enabled disk
|
465 |
templates
|
466 |
|
467 |
"""
|
468 |
assert constants.IPOLICY_DTS in ipolicy |
469 |
allowed_disk_templates = ipolicy[constants.IPOLICY_DTS] |
470 |
restricted_disk_templates = list(set(allowed_disk_templates) |
471 |
.intersection(set(enabled_disk_templates)))
|
472 |
ipolicy[constants.IPOLICY_DTS] = restricted_disk_templates |
473 |
|
474 |
|
475 |
def _InitCheckDrbdHelper(drbd_helper, drbd_enabled): |
476 |
"""Checks the DRBD usermode helper.
|
477 |
|
478 |
@type drbd_helper: string
|
479 |
@param drbd_helper: name of the DRBD usermode helper that the system should
|
480 |
use
|
481 |
|
482 |
"""
|
483 |
if not drbd_enabled: |
484 |
return
|
485 |
|
486 |
if drbd_helper is not None: |
487 |
try:
|
488 |
curr_helper = drbd.DRBD8.GetUsermodeHelper() |
489 |
except errors.BlockDeviceError, err:
|
490 |
raise errors.OpPrereqError("Error while checking drbd helper" |
491 |
" (disable drbd with --enabled-disk-templates"
|
492 |
" if you are not using drbd): %s" % str(err), |
493 |
errors.ECODE_ENVIRON) |
494 |
if drbd_helper != curr_helper:
|
495 |
raise errors.OpPrereqError("Error: requiring %s as drbd helper but %s" |
496 |
" is the current helper" % (drbd_helper,
|
497 |
curr_helper), |
498 |
errors.ECODE_INVAL) |
499 |
|
500 |
|
501 |
def InitCluster(cluster_name, mac_prefix, # pylint: disable=R0913, R0914 |
502 |
master_netmask, master_netdev, file_storage_dir, |
503 |
shared_file_storage_dir, candidate_pool_size, secondary_ip=None,
|
504 |
vg_name=None, beparams=None, nicparams=None, ndparams=None, |
505 |
hvparams=None, diskparams=None, enabled_hypervisors=None, |
506 |
modify_etc_hosts=True, modify_ssh_setup=True, |
507 |
maintain_node_health=False, drbd_helper=None, uid_pool=None, |
508 |
default_iallocator=None, primary_ip_version=None, ipolicy=None, |
509 |
prealloc_wipe_disks=False, use_external_mip_script=False, |
510 |
hv_state=None, disk_state=None, enabled_disk_templates=None): |
511 |
"""Initialise the cluster.
|
512 |
|
513 |
@type candidate_pool_size: int
|
514 |
@param candidate_pool_size: master candidate pool size
|
515 |
@type enabled_disk_templates: list of string
|
516 |
@param enabled_disk_templates: list of disk_templates to be used in this
|
517 |
cluster
|
518 |
|
519 |
"""
|
520 |
# TODO: complete the docstring
|
521 |
if config.ConfigWriter.IsCluster():
|
522 |
raise errors.OpPrereqError("Cluster is already initialised", |
523 |
errors.ECODE_STATE) |
524 |
|
525 |
if not enabled_hypervisors: |
526 |
raise errors.OpPrereqError("Enabled hypervisors list must contain at" |
527 |
" least one member", errors.ECODE_INVAL)
|
528 |
invalid_hvs = set(enabled_hypervisors) - constants.HYPER_TYPES
|
529 |
if invalid_hvs:
|
530 |
raise errors.OpPrereqError("Enabled hypervisors contains invalid" |
531 |
" entries: %s" % invalid_hvs,
|
532 |
errors.ECODE_INVAL) |
533 |
|
534 |
_InitCheckEnabledDiskTemplates(enabled_disk_templates) |
535 |
|
536 |
try:
|
537 |
ipcls = netutils.IPAddress.GetClassFromIpVersion(primary_ip_version) |
538 |
except errors.ProgrammerError:
|
539 |
raise errors.OpPrereqError("Invalid primary ip version: %d." % |
540 |
primary_ip_version, errors.ECODE_INVAL) |
541 |
|
542 |
hostname = netutils.GetHostname(family=ipcls.family) |
543 |
if not ipcls.IsValid(hostname.ip): |
544 |
raise errors.OpPrereqError("This host's IP (%s) is not a valid IPv%d" |
545 |
" address." % (hostname.ip, primary_ip_version),
|
546 |
errors.ECODE_INVAL) |
547 |
|
548 |
if ipcls.IsLoopback(hostname.ip):
|
549 |
raise errors.OpPrereqError("This host's IP (%s) resolves to a loopback" |
550 |
" address. Please fix DNS or %s." %
|
551 |
(hostname.ip, pathutils.ETC_HOSTS), |
552 |
errors.ECODE_ENVIRON) |
553 |
|
554 |
if not ipcls.Own(hostname.ip): |
555 |
raise errors.OpPrereqError("Inconsistency: this host's name resolves" |
556 |
" to %s,\nbut this ip address does not"
|
557 |
" belong to this host" %
|
558 |
hostname.ip, errors.ECODE_ENVIRON) |
559 |
|
560 |
clustername = netutils.GetHostname(name=cluster_name, family=ipcls.family) |
561 |
|
562 |
if netutils.TcpPing(clustername.ip, constants.DEFAULT_NODED_PORT, timeout=5): |
563 |
raise errors.OpPrereqError("Cluster IP already active", |
564 |
errors.ECODE_NOTUNIQUE) |
565 |
|
566 |
if not secondary_ip: |
567 |
if primary_ip_version == constants.IP6_VERSION:
|
568 |
raise errors.OpPrereqError("When using a IPv6 primary address, a valid" |
569 |
" IPv4 address must be given as secondary",
|
570 |
errors.ECODE_INVAL) |
571 |
secondary_ip = hostname.ip |
572 |
|
573 |
if not netutils.IP4Address.IsValid(secondary_ip): |
574 |
raise errors.OpPrereqError("Secondary IP address (%s) has to be a valid" |
575 |
" IPv4 address." % secondary_ip,
|
576 |
errors.ECODE_INVAL) |
577 |
|
578 |
if not netutils.IP4Address.Own(secondary_ip): |
579 |
raise errors.OpPrereqError("You gave %s as secondary IP," |
580 |
" but it does not belong to this host." %
|
581 |
secondary_ip, errors.ECODE_ENVIRON) |
582 |
|
583 |
if master_netmask is not None: |
584 |
if not ipcls.ValidateNetmask(master_netmask): |
585 |
raise errors.OpPrereqError("CIDR netmask (%s) not valid for IPv%s " % |
586 |
(master_netmask, primary_ip_version), |
587 |
errors.ECODE_INVAL) |
588 |
else:
|
589 |
master_netmask = ipcls.iplen |
590 |
|
591 |
if vg_name:
|
592 |
# Check if volume group is valid
|
593 |
vgstatus = utils.CheckVolumeGroupSize(utils.ListVolumeGroups(), vg_name, |
594 |
constants.MIN_VG_SIZE) |
595 |
if vgstatus:
|
596 |
raise errors.OpPrereqError("Error: %s" % vgstatus, errors.ECODE_INVAL) |
597 |
|
598 |
drbd_enabled = constants.DT_DRBD8 in enabled_disk_templates
|
599 |
_InitCheckDrbdHelper(drbd_helper, drbd_enabled) |
600 |
|
601 |
logging.debug("Stopping daemons (if any are running)")
|
602 |
result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-all"])
|
603 |
if result.failed:
|
604 |
raise errors.OpExecError("Could not stop daemons, command %s" |
605 |
" had exitcode %s and error '%s'" %
|
606 |
(result.cmd, result.exit_code, result.output)) |
607 |
|
608 |
file_storage_dir = _PrepareFileStorage(enabled_disk_templates, |
609 |
file_storage_dir) |
610 |
shared_file_storage_dir = _PrepareSharedFileStorage(enabled_disk_templates, |
611 |
shared_file_storage_dir) |
612 |
|
613 |
if not re.match("^[0-9a-z]{2}:[0-9a-z]{2}:[0-9a-z]{2}$", mac_prefix): |
614 |
raise errors.OpPrereqError("Invalid mac prefix given '%s'" % mac_prefix, |
615 |
errors.ECODE_INVAL) |
616 |
|
617 |
result = utils.RunCmd(["ip", "link", "show", "dev", master_netdev]) |
618 |
if result.failed:
|
619 |
raise errors.OpPrereqError("Invalid master netdev given (%s): '%s'" % |
620 |
(master_netdev, |
621 |
result.output.strip()), errors.ECODE_INVAL) |
622 |
|
623 |
dirs = [(pathutils.RUN_DIR, constants.RUN_DIRS_MODE)] |
624 |
utils.EnsureDirs(dirs) |
625 |
|
626 |
objects.UpgradeBeParams(beparams) |
627 |
utils.ForceDictType(beparams, constants.BES_PARAMETER_TYPES) |
628 |
utils.ForceDictType(nicparams, constants.NICS_PARAMETER_TYPES) |
629 |
|
630 |
objects.NIC.CheckParameterSyntax(nicparams) |
631 |
|
632 |
full_ipolicy = objects.FillIPolicy(constants.IPOLICY_DEFAULTS, ipolicy) |
633 |
_RestrictIpolicyToEnabledDiskTemplates(full_ipolicy, enabled_disk_templates) |
634 |
|
635 |
if ndparams is not None: |
636 |
utils.ForceDictType(ndparams, constants.NDS_PARAMETER_TYPES) |
637 |
else:
|
638 |
ndparams = dict(constants.NDC_DEFAULTS)
|
639 |
|
640 |
# This is ugly, as we modify the dict itself
|
641 |
# FIXME: Make utils.ForceDictType pure functional or write a wrapper
|
642 |
# around it
|
643 |
if hv_state:
|
644 |
for hvname, hvs_data in hv_state.items(): |
645 |
utils.ForceDictType(hvs_data, constants.HVSTS_PARAMETER_TYPES) |
646 |
hv_state[hvname] = objects.Cluster.SimpleFillHvState(hvs_data) |
647 |
else:
|
648 |
hv_state = dict((hvname, constants.HVST_DEFAULTS)
|
649 |
for hvname in enabled_hypervisors) |
650 |
|
651 |
# FIXME: disk_state has no default values yet
|
652 |
if disk_state:
|
653 |
for storage, ds_data in disk_state.items(): |
654 |
if storage not in constants.DS_VALID_TYPES: |
655 |
raise errors.OpPrereqError("Invalid storage type in disk state: %s" % |
656 |
storage, errors.ECODE_INVAL) |
657 |
for ds_name, state in ds_data.items(): |
658 |
utils.ForceDictType(state, constants.DSS_PARAMETER_TYPES) |
659 |
ds_data[ds_name] = objects.Cluster.SimpleFillDiskState(state) |
660 |
|
661 |
# hvparams is a mapping of hypervisor->hvparams dict
|
662 |
for hv_name, hv_params in hvparams.iteritems(): |
663 |
utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES) |
664 |
hv_class = hypervisor.GetHypervisor(hv_name) |
665 |
hv_class.CheckParameterSyntax(hv_params) |
666 |
|
667 |
# diskparams is a mapping of disk-template->diskparams dict
|
668 |
for template, dt_params in diskparams.items(): |
669 |
param_keys = set(dt_params.keys())
|
670 |
default_param_keys = set(constants.DISK_DT_DEFAULTS[template].keys())
|
671 |
if not (param_keys <= default_param_keys): |
672 |
unknown_params = param_keys - default_param_keys |
673 |
raise errors.OpPrereqError("Invalid parameters for disk template %s:" |
674 |
" %s" % (template,
|
675 |
utils.CommaJoin(unknown_params)), |
676 |
errors.ECODE_INVAL) |
677 |
utils.ForceDictType(dt_params, constants.DISK_DT_TYPES) |
678 |
if template == constants.DT_DRBD8 and vg_name is not None: |
679 |
# The default METAVG value is equal to the VG name set at init time,
|
680 |
# if provided
|
681 |
dt_params[constants.DRBD_DEFAULT_METAVG] = vg_name |
682 |
|
683 |
try:
|
684 |
utils.VerifyDictOptions(diskparams, constants.DISK_DT_DEFAULTS) |
685 |
except errors.OpPrereqError, err:
|
686 |
raise errors.OpPrereqError("While verify diskparam options: %s" % err, |
687 |
errors.ECODE_INVAL) |
688 |
|
689 |
# set up ssh config and /etc/hosts
|
690 |
rsa_sshkey = ""
|
691 |
dsa_sshkey = ""
|
692 |
if os.path.isfile(pathutils.SSH_HOST_RSA_PUB):
|
693 |
sshline = utils.ReadFile(pathutils.SSH_HOST_RSA_PUB) |
694 |
rsa_sshkey = sshline.split(" ")[1] |
695 |
if os.path.isfile(pathutils.SSH_HOST_DSA_PUB):
|
696 |
sshline = utils.ReadFile(pathutils.SSH_HOST_DSA_PUB) |
697 |
dsa_sshkey = sshline.split(" ")[1] |
698 |
if not rsa_sshkey and not dsa_sshkey: |
699 |
raise errors.OpPrereqError("Failed to find SSH public keys", |
700 |
errors.ECODE_ENVIRON) |
701 |
|
702 |
if modify_etc_hosts:
|
703 |
utils.AddHostToEtcHosts(hostname.name, hostname.ip) |
704 |
|
705 |
if modify_ssh_setup:
|
706 |
_InitSSHSetup() |
707 |
|
708 |
if default_iallocator is not None: |
709 |
alloc_script = utils.FindFile(default_iallocator, |
710 |
constants.IALLOCATOR_SEARCH_PATH, |
711 |
os.path.isfile) |
712 |
if alloc_script is None: |
713 |
raise errors.OpPrereqError("Invalid default iallocator script '%s'" |
714 |
" specified" % default_iallocator,
|
715 |
errors.ECODE_INVAL) |
716 |
elif constants.HTOOLS:
|
717 |
# htools was enabled at build-time, we default to it
|
718 |
if utils.FindFile(constants.IALLOC_HAIL,
|
719 |
constants.IALLOCATOR_SEARCH_PATH, |
720 |
os.path.isfile): |
721 |
default_iallocator = constants.IALLOC_HAIL |
722 |
|
723 |
now = time.time() |
724 |
|
725 |
# init of cluster config file
|
726 |
cluster_config = objects.Cluster( |
727 |
serial_no=1,
|
728 |
rsahostkeypub=rsa_sshkey, |
729 |
dsahostkeypub=dsa_sshkey, |
730 |
highest_used_port=(constants.FIRST_DRBD_PORT - 1),
|
731 |
mac_prefix=mac_prefix, |
732 |
volume_group_name=vg_name, |
733 |
tcpudp_port_pool=set(),
|
734 |
master_ip=clustername.ip, |
735 |
master_netmask=master_netmask, |
736 |
master_netdev=master_netdev, |
737 |
cluster_name=clustername.name, |
738 |
file_storage_dir=file_storage_dir, |
739 |
shared_file_storage_dir=shared_file_storage_dir, |
740 |
enabled_hypervisors=enabled_hypervisors, |
741 |
beparams={constants.PP_DEFAULT: beparams}, |
742 |
nicparams={constants.PP_DEFAULT: nicparams}, |
743 |
ndparams=ndparams, |
744 |
hvparams=hvparams, |
745 |
diskparams=diskparams, |
746 |
candidate_pool_size=candidate_pool_size, |
747 |
modify_etc_hosts=modify_etc_hosts, |
748 |
modify_ssh_setup=modify_ssh_setup, |
749 |
uid_pool=uid_pool, |
750 |
ctime=now, |
751 |
mtime=now, |
752 |
maintain_node_health=maintain_node_health, |
753 |
drbd_usermode_helper=drbd_helper, |
754 |
default_iallocator=default_iallocator, |
755 |
primary_ip_family=ipcls.family, |
756 |
prealloc_wipe_disks=prealloc_wipe_disks, |
757 |
use_external_mip_script=use_external_mip_script, |
758 |
ipolicy=full_ipolicy, |
759 |
hv_state_static=hv_state, |
760 |
disk_state_static=disk_state, |
761 |
enabled_disk_templates=enabled_disk_templates, |
762 |
) |
763 |
master_node_config = objects.Node(name=hostname.name, |
764 |
primary_ip=hostname.ip, |
765 |
secondary_ip=secondary_ip, |
766 |
serial_no=1,
|
767 |
master_candidate=True,
|
768 |
offline=False, drained=False, |
769 |
ctime=now, mtime=now, |
770 |
) |
771 |
InitConfig(constants.CONFIG_VERSION, cluster_config, master_node_config) |
772 |
cfg = config.ConfigWriter(offline=True)
|
773 |
ssh.WriteKnownHostsFile(cfg, pathutils.SSH_KNOWN_HOSTS_FILE) |
774 |
cfg.Update(cfg.GetClusterInfo(), logging.error) |
775 |
ssconf.WriteSsconfFiles(cfg.GetSsconfValues()) |
776 |
|
777 |
# set up the inter-node password and certificate
|
778 |
_InitGanetiServerSetup(hostname.name) |
779 |
|
780 |
logging.debug("Starting daemons")
|
781 |
result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-all"])
|
782 |
if result.failed:
|
783 |
raise errors.OpExecError("Could not start daemons, command %s" |
784 |
" had exitcode %s and error %s" %
|
785 |
(result.cmd, result.exit_code, result.output)) |
786 |
|
787 |
_WaitForMasterDaemon() |
788 |
|
789 |
|
790 |
def InitConfig(version, cluster_config, master_node_config, |
791 |
cfg_file=pathutils.CLUSTER_CONF_FILE): |
792 |
"""Create the initial cluster configuration.
|
793 |
|
794 |
It will contain the current node, which will also be the master
|
795 |
node, and no instances.
|
796 |
|
797 |
@type version: int
|
798 |
@param version: configuration version
|
799 |
@type cluster_config: L{objects.Cluster}
|
800 |
@param cluster_config: cluster configuration
|
801 |
@type master_node_config: L{objects.Node}
|
802 |
@param master_node_config: master node configuration
|
803 |
@type cfg_file: string
|
804 |
@param cfg_file: configuration file path
|
805 |
|
806 |
"""
|
807 |
uuid_generator = config.TemporaryReservationManager() |
808 |
cluster_config.uuid = uuid_generator.Generate([], utils.NewUUID, |
809 |
_INITCONF_ECID) |
810 |
master_node_config.uuid = uuid_generator.Generate([], utils.NewUUID, |
811 |
_INITCONF_ECID) |
812 |
cluster_config.master_node = master_node_config.uuid |
813 |
nodes = { |
814 |
master_node_config.uuid: master_node_config, |
815 |
} |
816 |
default_nodegroup = objects.NodeGroup( |
817 |
uuid=uuid_generator.Generate([], utils.NewUUID, _INITCONF_ECID), |
818 |
name=constants.INITIAL_NODE_GROUP_NAME, |
819 |
members=[master_node_config.uuid], |
820 |
diskparams={}, |
821 |
) |
822 |
nodegroups = { |
823 |
default_nodegroup.uuid: default_nodegroup, |
824 |
} |
825 |
now = time.time() |
826 |
config_data = objects.ConfigData(version=version, |
827 |
cluster=cluster_config, |
828 |
nodegroups=nodegroups, |
829 |
nodes=nodes, |
830 |
instances={}, |
831 |
networks={}, |
832 |
serial_no=1,
|
833 |
ctime=now, mtime=now) |
834 |
utils.WriteFile(cfg_file, |
835 |
data=serializer.Dump(config_data.ToDict()), |
836 |
mode=0600)
|
837 |
|
838 |
|
839 |
def FinalizeClusterDestroy(master_uuid): |
840 |
"""Execute the last steps of cluster destroy
|
841 |
|
842 |
This function shuts down all the daemons, completing the destroy
|
843 |
begun in cmdlib.LUDestroyOpcode.
|
844 |
|
845 |
"""
|
846 |
cfg = config.ConfigWriter() |
847 |
modify_ssh_setup = cfg.GetClusterInfo().modify_ssh_setup |
848 |
runner = rpc.BootstrapRunner() |
849 |
|
850 |
master_name = cfg.GetNodeName(master_uuid) |
851 |
|
852 |
master_params = cfg.GetMasterNetworkParameters() |
853 |
master_params.uuid = master_uuid |
854 |
ems = cfg.GetUseExternalMipScript() |
855 |
result = runner.call_node_deactivate_master_ip(master_name, master_params, |
856 |
ems) |
857 |
|
858 |
msg = result.fail_msg |
859 |
if msg:
|
860 |
logging.warning("Could not disable the master IP: %s", msg)
|
861 |
|
862 |
result = runner.call_node_stop_master(master_name) |
863 |
msg = result.fail_msg |
864 |
if msg:
|
865 |
logging.warning("Could not disable the master role: %s", msg)
|
866 |
|
867 |
result = runner.call_node_leave_cluster(master_name, modify_ssh_setup) |
868 |
msg = result.fail_msg |
869 |
if msg:
|
870 |
logging.warning("Could not shutdown the node daemon and cleanup"
|
871 |
" the node: %s", msg)
|
872 |
|
873 |
|
874 |
def SetupNodeDaemon(opts, cluster_name, node): |
875 |
"""Add a node to the cluster.
|
876 |
|
877 |
This function must be called before the actual opcode, and will ssh
|
878 |
to the remote node, copy the needed files, and start ganeti-noded,
|
879 |
allowing the master to do the rest via normal rpc calls.
|
880 |
|
881 |
@param cluster_name: the cluster name
|
882 |
@param node: the name of the new node
|
883 |
|
884 |
"""
|
885 |
data = { |
886 |
constants.NDS_CLUSTER_NAME: cluster_name, |
887 |
constants.NDS_NODE_DAEMON_CERTIFICATE: |
888 |
utils.ReadFile(pathutils.NODED_CERT_FILE), |
889 |
constants.NDS_SSCONF: ssconf.SimpleStore().ReadAll(), |
890 |
constants.NDS_START_NODE_DAEMON: True,
|
891 |
} |
892 |
|
893 |
RunNodeSetupCmd(cluster_name, node, pathutils.NODE_DAEMON_SETUP, |
894 |
opts.debug, opts.verbose, |
895 |
True, opts.ssh_key_check, opts.ssh_key_check, data)
|
896 |
|
897 |
_WaitForNodeDaemon(node) |
898 |
|
899 |
|
900 |
def MasterFailover(no_voting=False): |
901 |
"""Failover the master node.
|
902 |
|
903 |
This checks that we are not already the master, and will cause the
|
904 |
current master to cease being master, and the non-master to become
|
905 |
new master.
|
906 |
|
907 |
@type no_voting: boolean
|
908 |
@param no_voting: force the operation without remote nodes agreement
|
909 |
(dangerous)
|
910 |
|
911 |
"""
|
912 |
sstore = ssconf.SimpleStore() |
913 |
|
914 |
old_master, new_master = ssconf.GetMasterAndMyself(sstore) |
915 |
node_names = sstore.GetNodeList() |
916 |
mc_list = sstore.GetMasterCandidates() |
917 |
|
918 |
if old_master == new_master:
|
919 |
raise errors.OpPrereqError("This commands must be run on the node" |
920 |
" where you want the new master to be."
|
921 |
" %s is already the master" %
|
922 |
old_master, errors.ECODE_INVAL) |
923 |
|
924 |
if new_master not in mc_list: |
925 |
mc_no_master = [name for name in mc_list if name != old_master] |
926 |
raise errors.OpPrereqError("This node is not among the nodes marked" |
927 |
" as master candidates. Only these nodes"
|
928 |
" can become masters. Current list of"
|
929 |
" master candidates is:\n"
|
930 |
"%s" % ("\n".join(mc_no_master)), |
931 |
errors.ECODE_STATE) |
932 |
|
933 |
if not no_voting: |
934 |
vote_list = GatherMasterVotes(node_names) |
935 |
|
936 |
if vote_list:
|
937 |
voted_master = vote_list[0][0] |
938 |
if voted_master is None: |
939 |
raise errors.OpPrereqError("Cluster is inconsistent, most nodes did" |
940 |
" not respond.", errors.ECODE_ENVIRON)
|
941 |
elif voted_master != old_master:
|
942 |
raise errors.OpPrereqError("I have a wrong configuration, I believe" |
943 |
" the master is %s but the other nodes"
|
944 |
" voted %s. Please resync the configuration"
|
945 |
" of this node." %
|
946 |
(old_master, voted_master), |
947 |
errors.ECODE_STATE) |
948 |
# end checks
|
949 |
|
950 |
rcode = 0
|
951 |
|
952 |
logging.info("Setting master to %s, old master: %s", new_master, old_master)
|
953 |
|
954 |
try:
|
955 |
# instantiate a real config writer, as we now know we have the
|
956 |
# configuration data
|
957 |
cfg = config.ConfigWriter(accept_foreign=True)
|
958 |
|
959 |
old_master_node = cfg.GetNodeInfoByName(old_master) |
960 |
if old_master_node is None: |
961 |
raise errors.OpPrereqError("Could not find old master node '%s' in" |
962 |
" cluster configuration." % old_master,
|
963 |
errors.ECODE_NOENT) |
964 |
|
965 |
cluster_info = cfg.GetClusterInfo() |
966 |
new_master_node = cfg.GetNodeInfoByName(new_master) |
967 |
if new_master_node is None: |
968 |
raise errors.OpPrereqError("Could not find new master node '%s' in" |
969 |
" cluster configuration." % new_master,
|
970 |
errors.ECODE_NOENT) |
971 |
|
972 |
cluster_info.master_node = new_master_node.uuid |
973 |
# this will also regenerate the ssconf files, since we updated the
|
974 |
# cluster info
|
975 |
cfg.Update(cluster_info, logging.error) |
976 |
except errors.ConfigurationError, err:
|
977 |
logging.error("Error while trying to set the new master: %s",
|
978 |
str(err))
|
979 |
return 1 |
980 |
|
981 |
# if cfg.Update worked, then it means the old master daemon won't be
|
982 |
# able now to write its own config file (we rely on locking in both
|
983 |
# backend.UploadFile() and ConfigWriter._Write(); hence the next
|
984 |
# step is to kill the old master
|
985 |
|
986 |
logging.info("Stopping the master daemon on node %s", old_master)
|
987 |
|
988 |
runner = rpc.BootstrapRunner() |
989 |
master_params = cfg.GetMasterNetworkParameters() |
990 |
master_params.uuid = old_master_node.uuid |
991 |
ems = cfg.GetUseExternalMipScript() |
992 |
result = runner.call_node_deactivate_master_ip(old_master, |
993 |
master_params, ems) |
994 |
|
995 |
msg = result.fail_msg |
996 |
if msg:
|
997 |
logging.warning("Could not disable the master IP: %s", msg)
|
998 |
|
999 |
result = runner.call_node_stop_master(old_master) |
1000 |
msg = result.fail_msg |
1001 |
if msg:
|
1002 |
logging.error("Could not disable the master role on the old master"
|
1003 |
" %s, please disable manually: %s", old_master, msg)
|
1004 |
|
1005 |
logging.info("Checking master IP non-reachability...")
|
1006 |
|
1007 |
master_ip = sstore.GetMasterIP() |
1008 |
total_timeout = 30
|
1009 |
|
1010 |
# Here we have a phase where no master should be running
|
1011 |
def _check_ip(): |
1012 |
if netutils.TcpPing(master_ip, constants.DEFAULT_NODED_PORT):
|
1013 |
raise utils.RetryAgain()
|
1014 |
|
1015 |
try:
|
1016 |
utils.Retry(_check_ip, (1, 1.5, 5), total_timeout) |
1017 |
except utils.RetryTimeout:
|
1018 |
logging.warning("The master IP is still reachable after %s seconds,"
|
1019 |
" continuing but activating the master on the current"
|
1020 |
" node will probably fail", total_timeout)
|
1021 |
|
1022 |
if jstore.CheckDrainFlag():
|
1023 |
logging.info("Undraining job queue")
|
1024 |
jstore.SetDrainFlag(False)
|
1025 |
|
1026 |
logging.info("Starting the master daemons on the new master")
|
1027 |
|
1028 |
result = rpc.BootstrapRunner().call_node_start_master_daemons(new_master, |
1029 |
no_voting) |
1030 |
msg = result.fail_msg |
1031 |
if msg:
|
1032 |
logging.error("Could not start the master role on the new master"
|
1033 |
" %s, please check: %s", new_master, msg)
|
1034 |
rcode = 1
|
1035 |
|
1036 |
logging.info("Master failed over from %s to %s", old_master, new_master)
|
1037 |
return rcode
|
1038 |
|
1039 |
|
1040 |
def GetMaster(): |
1041 |
"""Returns the current master node.
|
1042 |
|
1043 |
This is a separate function in bootstrap since it's needed by
|
1044 |
gnt-cluster, and instead of importing directly ssconf, it's better
|
1045 |
to abstract it in bootstrap, where we do use ssconf in other
|
1046 |
functions too.
|
1047 |
|
1048 |
"""
|
1049 |
sstore = ssconf.SimpleStore() |
1050 |
|
1051 |
old_master, _ = ssconf.GetMasterAndMyself(sstore) |
1052 |
|
1053 |
return old_master
|
1054 |
|
1055 |
|
1056 |
def GatherMasterVotes(node_names): |
1057 |
"""Check the agreement on who is the master.
|
1058 |
|
1059 |
This function will return a list of (node, number of votes), ordered
|
1060 |
by the number of votes. Errors will be denoted by the key 'None'.
|
1061 |
|
1062 |
Note that the sum of votes is the number of nodes this machine
|
1063 |
knows, whereas the number of entries in the list could be different
|
1064 |
(if some nodes vote for another master).
|
1065 |
|
1066 |
We remove ourselves from the list since we know that (bugs aside)
|
1067 |
since we use the same source for configuration information for both
|
1068 |
backend and boostrap, we'll always vote for ourselves.
|
1069 |
|
1070 |
@type node_names: list
|
1071 |
@param node_names: the list of nodes to query for master info; the current
|
1072 |
node will be removed if it is in the list
|
1073 |
@rtype: list
|
1074 |
@return: list of (node, votes)
|
1075 |
|
1076 |
"""
|
1077 |
myself = netutils.Hostname.GetSysName() |
1078 |
try:
|
1079 |
node_names.remove(myself) |
1080 |
except ValueError: |
1081 |
pass
|
1082 |
if not node_names: |
1083 |
# no nodes left (eventually after removing myself)
|
1084 |
return []
|
1085 |
results = rpc.BootstrapRunner().call_master_info(node_names) |
1086 |
if not isinstance(results, dict): |
1087 |
# this should not happen (unless internal error in rpc)
|
1088 |
logging.critical("Can't complete rpc call, aborting master startup")
|
1089 |
return [(None, len(node_names))] |
1090 |
votes = {} |
1091 |
for node_name in results: |
1092 |
nres = results[node_name] |
1093 |
data = nres.payload |
1094 |
msg = nres.fail_msg |
1095 |
fail = False
|
1096 |
if msg:
|
1097 |
logging.warning("Error contacting node %s: %s", node_name, msg)
|
1098 |
fail = True
|
1099 |
# for now we accept both length 3, 4 and 5 (data[3] is primary ip version
|
1100 |
# and data[4] is the master netmask)
|
1101 |
elif not isinstance(data, (tuple, list)) or len(data) < 3: |
1102 |
logging.warning("Invalid data received from node %s: %s",
|
1103 |
node_name, data) |
1104 |
fail = True
|
1105 |
if fail:
|
1106 |
if None not in votes: |
1107 |
votes[None] = 0 |
1108 |
votes[None] += 1 |
1109 |
continue
|
1110 |
master_node = data[2]
|
1111 |
if master_node not in votes: |
1112 |
votes[master_node] = 0
|
1113 |
votes[master_node] += 1
|
1114 |
|
1115 |
vote_list = [v for v in votes.items()] |
1116 |
# sort first on number of votes then on name, since we want None
|
1117 |
# sorted later if we have the half of the nodes not responding, and
|
1118 |
# half voting all for the same master
|
1119 |
vote_list.sort(key=lambda x: (x[1], x[0]), reverse=True) |
1120 |
|
1121 |
return vote_list
|